Imaging and analysis of isomeric unsaturated lipids through online photochemical derivatization of C=C bonds

Abstract

Unraveling the complexity of the lipidome requires the development of novel approaches for the structural characterization of lipid species with isomer-level discrimination. Herein, we introduce an online photochemical approach for lipid isomer identification through selective derivatization of double bonds by reaction with singlet oxygen. Lipid hydroperoxide products are generated promptly after laser irradiation. Fragmentation of these species in a mass spectrometer produces diagnostic fragments, which reveal the C=C locations in the unreacted lipids. This approach uses an inexpensive light source and photosensitizer making it easy to incorporate into any lipidomics workflow. We demonstrate the utility of this approach for the shotgun profiling of C=C locations in different lipid classes present in tissue extracts using electrospray ionization (ESI) and ambient imaging of lipid species differing only by the location of C=C bonds using nanospray desorption electrospray ionization (nano-DESI).

Graphical Abstract

Rapid derivatization of lipids using a photoinitiated reaction with singlet oxygen provides insights into the C=C bond location in the precursor lipid through diagnostic CID fragments. This simple and inexpensive method is compatible with lipidomics workflows. It can be adapted for the analysis tissue extracts using electrospray ionization (ESI) and for imaging of isomeric lipids in tissue sections using nanospray desorption electrospray ionization (nano-DESI).

Lipids are essential biomolecules acting as structural components of membranes, energy reservoirs, or signaling molecules in biological systems.^[1] Multiple biochemical transformations that lipids undergo during their biosynthesis generate diverse lipid structures, which complicates their structural characterization.^[2] Mass spectrometry (MS) has been extensively used to identify lipid classes, acyl chain composition, and degree of unsaturation.^[3] However, the differentiation of isomers that vary only by the position of C=C double bonds critical to understanding their role in key metabolic processes remains challenging. Recent studies have attributed changes in the relative abundance of positional isomers to altered lipid metabolism in cancer^[4,5] and diabetes.^[6] These findings inspire the development of new approaches for the identification of C=C bond locations in unsaturated lipids.^[7]

The most successful MS-based strategies for identifying C=C bond positions include ozone-induced dissociation,^[8,9] ion-ion reactions,^[10] electron-ion reaction-based dissociation,^[11,12] and ultraviolet photodissociation.^[13,14] Alternatively, derivatization methods including Paternò Büchi,^[15,16] ozonolysis,^[17,18] epoxidation,^[4,19,20] and thiol-ene^[21] reactions have been used for the untargeted analysis of the double bond in unsaturated lipids thereby dramatically expanding the depth of molecular coverage obtained in lipidomics experiments. When coupled with mass spectrometry imaging (MSI), these approaches reveal the spatial localization of isomeric lipids in biological systems otherwise invisible with traditional MSI techniques. ^[5,22–24]

Despite the significant progress in this field, spatially-resolved isomer-selected lipidomics experiments are still challenging. Herein, we introduce an online derivatization approach, that 1) enables fast photooxidation and profiling of C=C locations in lipids, 2) uses an inexpensive visible light source which can be easily focused, 3) does not require any instrument modification, and 4) is compatible with both the electrospray ionization (ESI)-based shotgun lipidomics workflow for the analysis of lipid extracts and mass spectrometry imaging (MSI) of biological samples using nanospray desorption electrospray ionization (nano-DESI).

Nano-DESI is an ambient ionization technique, in which analytes are directly extracted from a sample into a solvent and ionized by ESI at a mass spectrometer inlet.^[25] The addition of reagents to the extraction solvent may be used to perform online chemical derivatization of analytes,^[26] which often benefits from reaction acceleration in microdroplets.^[27] Coupling of nano-DESI with the on-line photochemical reaction is a first step towards the development of derivatization approaches for isomer-selective imaging of lipids and metabolites in biological samples. The approach reported herein utilizes a reaction between singlet oxygen (¹O₂) with C=C bonds and enables efficient derivatization of unsaturated lipids on a timescale compatible with imaging experiments.

¹O₂ is the excited state of molecular oxygen containing one empty orbital, which reacts with electron-rich double bonds to produce hydroperoxides (Figure 1a).^[28,29] Photochemical generation of ¹O₂ is achieved using a photosensitizer, which is excited to a triplet state upon exposure to light and subsequently transfers its excitation energy to molecular oxygen present in solution. ¹O₂ reacts with lipids to form lipid hydroperoxides (LOOHs).^[30–32] When subjected to collision-induced dissociation (CID), LOOHs undergo a bond cleavage at the location of the hydroperoxide group (Scheme S1) thereby revealing the double bond position. Herein, we use the online reaction of unsaturated lipids with ¹O₂ followed by CID of the resulting LOOH to obtain diagnostic peaks that provide insights into the C=C bond location.

Figure 1.

a) Schematic representation of the oxidation of the C=C bonds in lipids by singlet oxygen and fragmentation via CID that yields unique neutral losses. Experimental setup for the online singlet oxygen reaction with lipids coupled to b) ESI and c) nano-DESI.

ESI experiments(Figure 1b) are described in detail in the supporting information (SI). In this experiment, a mixture of MeOH:H₂O (9/1, v/v) containing rose bengal (RB) as a photosensitizer and lipids is propelled through a fused silica capillary at 0.5 μL/min, ionized by ESI, and analyzed using MS. For the online derivatization, the capillary tip pulled to ~25 μm OD is illuminated with a green laser pointer (532 nm) positioned 15 cm away from the tip and focused using a convex lens. This configuration, which can be readily adapted to lipidomics experiments was used to test the feasibility of the online derivatization of C=C bonds in fatty acyl chains using the ¹O₂ reaction.

In another experiment (Figures 1c, S1), the ¹O₂ reaction is coupled to a high-resolution nano-DESI probe to perform an online derivatization and C=C profiling of analytes in complex lipid mixtures extracted from tissue sections. Here, the solvent containing RB is propelled through the nano-DESI probe, which extracts analytes from a specific location on the sample into a dynamic liquid bridge shown in the inset. The spray capillary transfers the extracted lipids and metabolites to a mass spectrometer inlet. The laser light is focused onto the tip of the spray capillary. A shear force probe is placed next to the nano-DESI probe to maintain a constant distance to the sample.^[33–35]

The reaction between ¹O₂ and unsaturated lipids (Figure 1a) generates an allylic hydroperoxide at either end of the double bond accompanied by a shift in the double bond position to the adjacent carbon.^[31,32] The reaction results in a mass shift of 32 Da corresponding to the net addition of O₂. Lipid nomenclature is as follows: LPE 17:1(10Δ) denotes a 17-carbon fatty acid with one degree of unsaturation and the C=C bond on the 10th carbon atom from the carboxylic end (10Δ). For lipid standards with known geometry, Δ is replaced with either Z (cis) or E (trans). We selected RB as a photosensitizer due to its high quantum yield.^[36]

Figure 2a shows ESI-MS of a standard mixture containing LPE 17:1, PC 36:2, and TG 48:3 obtained upon irradiation. Although the dominant product corresponds to the addition of one O₂, other LOOH products observed in the spectrum confirm that the reaction occurs at each double bond. No other side products such as aldehydes, alcohols or ketones were observed in these experiments (Figures S2–S3). The absence of side products commonly observed in the bulk experiments may be attributed to the lack of direct interaction between RB and lipids in the relatively dilute solutions analyzed in this study.

Figure 2.

Elucidation of the C=C location in positional isomeric standards. a) Positive mode ESI-MS of LPE 17:1, PC 36:2 and TG 48:3 and their (LOOH)_n products generated by the reaction with ¹O₂. The [M+Na]⁺ ions of (LOOH)_n where n indicates the number of O₂ additions are highlighted in blue. b) Structures of PC 18:1(9Z)/18:1(9Z) (left) and PC 18:1(6Z)/18:1(6Z) (right) showing the cleavage sites. c) LTQ-CID and d) HCD spectra of the isomeric LOOH products at m/z 840.5.

CID of LOOHs generates unique neutral losses (NL) directly from the precursor ion, which reveal the position of the hydroperoxide group and hence C=C bond in the unreacted lipid. This is illustrated using two isomeric LOOH species generated from two positional isomers: PC 18:1(9Z)/18:1(9Z) and PC 18:1(6Z)/18:1(6Z) (Figure 2b). In this experiment, 1 μM solutions of each standard containing 50 μM RB were analyzed individually using ESI coupled with laser excitation. For both isomers, CID spectra (Figures 2c–d) contain fragments at m/z 781.4 and 657.5 corresponding to head group losses from m/z 840.5. Meanwhile, fragmentation at the hydroperoxide generates different fragments for the two isomers. For the 9Z isomer, a primary fragment ion corresponding to the loss of C₈H₁₈O due to the cleavage at C10 of precursor A (Figure 1a) is observed at m/z 710.4. This fragment undergoes subsequent NL of 59 and 183 generating peaks at m/z 651.3 and 527.3, respectively confirming that LOOH is a PC species. The corresponding loss of C₁₁H₂₄O due to cleavage at C7 is observed for the 6Z isomer generating a fragment at m/z 668.4 along with the head group losses at m/z 609.3 and 485.3. Fragment ions corresponding to precursor B (Figure 1a) are not observed in the higher energy collision-induced dissociation (HCD) spectra but are observed as minor peaks in the LTQ-CID spectrum (Figure 2c). This effect is more pronounced for polyunsaturated LOOHs (Figures S4–S5), for which the diagnostic fragments of product B are observed in LTQ-CID but not in HCD spectra. We propose that differences in the energetics and kinetics of fragmentation of the two LOOH products affect their branching ratio on different MS timescales and levels of internal excitation. Importantly, a fragment corresponding to product A is observed on all the instruments and can be used to identify the C=C location of multiple lipid classes as shown in Figures S4–S8.

. Figure S9 shows mass spectra obtained for a mouse gastrocnemius muscle tissue extract containing different concentrations of RB (5, 50, 100, and 200 μM).The 50 μM RB solution provided good yields of LOOH products without producing dominant RB-related peaks that suppress the ionization of lipids at higher RB concentrations. Figure 3a shows part of a mass spectrum obtained under the optimized conditions, in which the LOOH products exhibit a 32 Da mass shift from their precursor lipids. Ion chronograms of the endogenous PC 38:6 at m/z 828.6 and PC 36:4 at m/z 804.6 and the corresponding LOOH products are depicted in Figure 3b. The signals of PC 38:6(LOOH) at m/z 860.5 and PC 36:4(LOOH) at m/z 836.5 appear once the light is turned on showing a 30–40% LOOH yield relative to the unreacted lipid. LOOH signals remain stable over time demonstrating the feasibility of this approach for the online chemical derivatization of C=C bonds. As described in the SI, we estimate that the reaction time is shorter than 250 ms.

Figure 3.

a) ESI-MS of a mouse gastrocnemius muscle tissue extract obtained using the ¹O₂ reaction; the LOOH products are shown in blue. b) Ion chronograms of the endogenous PC 38:6 at m/z 828.55 (left) and PC 36:4 m/z 804.55 (right) and their corresponding reaction products at m/z 860.54 and m/z 836.54, respectively, obtained with and without laser excitation.

The rapid and efficient LOOH formation makes the ¹O₂ reaction compatible with nano-DESI MSI experiments. To demonstrate the feasibility of isomer-specific profiling and imaging of lipids in tissues, we coupled the ¹O₂ reaction with a high-resolution nano-DESI probe. The details of these experiments conducted in both MS¹ (MS¹I) and MS² (MS²I) modes are provided in the SI. We used MeOH:H₂O (9/1, v/v) containing 50 μM RB and 200 nM LPE 17:1 as the extraction solvent. Figure S10 demonstrates the rapid response of the LPE 17:1 LOOH formation in the nano-DESI probe to laser excitation, which is similar to that observed using ESI.

Line profiles obtained for the LPE 17:1 standard, an endogenous PC 34:1, and their LOOH products by scanning the nano-DESI probe over a mouse uterine tissue are shown in Figure S11. Only LPE 17:1 is observed when the probe is positioned on the glass slide. After the light is turned on, the ion signal of LPE 17:1(LOOH) appears while the signal of LPE 17:1 decreases. The signals of the endogenous PC 34:1 and its product, PC 34:1(LOOH), are observed when the nano-DESI probe is on the tissue indicating that PC 34:1(LOOH) is generated with substantial yield on a timescale compatible with nano-DESI MSI. When the light is turned off, the signals of the reaction products LPE 17:1(LOOH) and PC 34:1(LOOH) disappear. Meanwhile, the signals of the corresponding precursor lipids remain. When the nano-DESI probe gets off the tissue, the signal of PC 34:1 disappears. This experiment establishes the utility of the ¹O₂ reaction for imaging experiments.

We identified several isomeric lipids sampled directly from tissues by the nano-DESI probe using CID on a targeted list of LOOH products. Tables S1 and S2 report unsaturated lipid species in mouse muscle and uterine tissues, for which C=C locations were successfully identified. Lipids containing FA 18:1 showed diagnostic fragments indicative of the presence of isomeric 9Δ and 11Δ pairs (Figure S12). Due to the complexity of the lipidome, some acyl chain combinations may yield ambiguous identifications in MS². For example, the same neutral losses are expected for FA 16:1(Δ7) and FA 18:1(Δ9) or for FA 16:1(Δ9) and FA 18:1(Δ11). Therefore, unambiguous identification of isomeric lipids containing both FA 16:1 and FA 18:1 requires MS³. Furthermore, some LOOHs overlap with endogenous lipids. Nevertheless, due to a very specific fragmentation, a majority of LOOHs and hence positional isomers are readily identified using MS².

MS²I experiments focused on the localization of 9Δ and 11Δ isomersin rat brain and mouse uterine tissue. Figure 4a shows a representative HCD spectrum of m/z 814.5 ± 1 in rat brain tissue. The spectrum contains product ions of different phospholipids present in this m/z window. PC 34:1(LOOH) at m/z 814.5573 shows product ions corresponding to 9Δ (684.4, 625.3 and 501.3) and 11Δ (712.5, 653.4 and 529.4) isomers. We also observed fragment ions of the isobaric species including the [M+Na]⁺ ion of PE 40:6 (m/z 814.5362) and [M+H]⁺ ion of PE 42:9 (m/z 814.5387) at m/z 771.5 and 673.5, respectively. Although the mass shift of 32 Da generated by the ¹O₂ reaction is rather small, which complicates MS¹ spectra, product ions of different isomeric and isobaric species are readily distinguished in the MS² spectrum.

Figure 4.

a) CID spectra of PC 34:1(LOOH) at m/z 814.5573 showing the presence of 9Δ and 11Δ isomers labeled in pink and blue, respectively. Fragment ions of the isobaric PE 40:6 (Na⁺) and PE 42:9 (H⁺) are labeled in green and yellow, respectively. b) Optical image of rat cerebellum. c) Nano-DESI MS¹I of PC 34:1(LOOH). FDIs of positional isomers corresponding to d) 11Δ/(9Δ+11Δ) and e) 9Δ/(9Δ+11Δ). Nano-DESI MS² images of the fragments at f) m/z 771.5 and g) m/z 673.5 corresponding to PE 40:6 and PE 42:9, respectively. The spatial resolution of 30 μm was determined as described in Figure S15.

Next, we examined the spatial localization of the isomeric species using fractional distribution images (FDI). FDIs were generated by plotting the ratio of the summed intensities of all the fragments of the 11Δ isomer to the summed intensities of all the fragments of 11Δ and 9Δ isomers as a function of location on the tissue. Figure 4b shows an optical image of the rat tissue analyzed in MSI experiments. An MS¹ image of PC 34:1(LOOH) (Figure 4c) indicates that this molecule is enhanced in the gray matter. In contrast, FDI of 11Δ (Figure 4c) indicates the lower abundance of 11Δ in comparison with the 9Δ isomer in the white matter. A complementary distribution of the 9Δ isomer is shown in Figure 4d. These results are consistent with previous studies.^[5,22–24] MS²I of the characteristic fragments of the isobaric PE 40:6 at m/z 771.5 and PE 42:9 at m/z 673.5 present in the same isolation window are shown in Figures 4f and 4g, respectively. We note that it is impossible to identify C=C bond positions in PE 40:6 and PE 42:9 because CID of non-oxidized species does not provide this information. The results shown in Figure 4 indicate that both species are enhanced in the gray matter. FDI of the 11Δ isomer of PC 18:0_18:1(LOOH) at m/z 842.5881 obtained in the same experiment is shown in Figure S13b. Meanwhile, MS²I of mouse uterine tissue indicates that PC 16:0_18:1(LOOH) is evenly distributed across the tissue (Figure S14). These results indicate that the localization of isomeric lipids is strongly dependent on tissue type.

In summary, we have developed an online chemical derivatization method based on the ¹O₂ reaction for identifying C=C positions in different lipid classes. Rapid reaction rates enable on-the-fly photooxidation of lipids into their corresponding LOOH products. CID of LOOHs produces unique fragments revealing the C=C bond positions. This approach relies on an inexpensive light source and photosensitizer added to the solvent or analyte mixture and can be readily implemented on any mass spectrometer. As a result, it is compatible both with lipidomics workflows and liquid extraction-based ambient ionization techniques. The small mass shift of 32 Da between the reactants and products may complicate MS¹ spectra but does not affect MS² experiments. We also note that RB generates abundant peaks in negative mode, which interfere with analyte signals. A different photosensitizer may be used in negative mode to alleviate this problem. Coupling of the ¹O₂ reaction with nano-DESI MS²I enables spatial localization of positional isomers, which cannot be achieved in the MS¹I mode. This capability will facilitate understanding of the role of isomeric C=C lipids in biological processes and exploit their potential as lipid biomarkers.